Computers and Structures 79 (2001) 243±252

www.elsevier.com/locate/compstruc

The inf±sup condition and its evaluation for mixed ®nite

element methods
Klaus-J
urgen Bathe *
Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Mass Avenue, Cambridge, MA 02139, USA
Accepted 16 November 1999

Abstract
The objective of this paper is to review the general inf±sup condition for mixed ®nite element methods and sum-
marize numerical procedures (inf±sup tests) for the evaluation of the inf±sup expressions speci®c to various problem
areas. The inf±sup testing of a given mixed ®nite element discretization is most important in order to assess its reliability
and solution e€ectiveness. The problem areas considered are (almost) incompressible analysis of solids and ¯uids,
acoustic ¯uids, the analysis of plates and shells, and the solution of convection-dominated ¯ows. Ó 2000 Elsevier
Science Ltd. All rights reserved.

1. Introduction already involve multiple ®eld variables; for example, for

The ®nite element method is now abundantly used
for the analysis of solids and structures, ¯uids and ¯uid-
structure interactions. For the analysis of solids and
structures, displacement-based ®nite element discretiza-
tions are most attractive; however, for some analyses,
mixed methods need to be employed. For the analyses of
¯uid ¯ows, mixed ®nite element procedures are generally
used [1].
While ± for the analysis of solids and structures ± the
displacement-based ®nite element procedures are for-
mulated in a straightforward manner using the principle
of virtual displacements, e€ective mixed methods are
much more dicult to establish. These formulations
must be employed for the analyses of (almost) incom-
pressible response, such as those encountered in elastic
solutions of rubber-like materials, and in inelastic solu-
tions of many materials. Mixed formulations need also
be employed for the analyses of plates and shells.
Considering ¯uid ¯ows, it is natural to use a mixed
formulation because the governing di€erential equations
*
Tel.: +1-617-2536645; fax: +1-617-2532275.
incompressible ¯uid ¯ow, the velocities and pressure are
natural solution variables to use. For the analysis of
¯uid-structure interactions, therefore, mixed formula-
tions are also widely used.
The ®rst mixed formulations were proposed using
suitable variational principles of the continuum prob-
lem, such as the Hu±Washizu and Hellinger±Reissner
principles. Reasonable ®nite element interpolations were
chosen and numerical test cases were solved to demon-
strate the e€ectiveness of the formulations. However,
while for the displacement-based ®nite element method
a few test cases can indeed show the overall e€ective-
ness of a discretization scheme, a mixed method
may work remarkably well in the solution of some
problems and totally fail in other problems. The fact
that the mixed ®nite element method is based on a
proper variational formulation does not at all ensure
that the ®nite element method is reliable and that the
method can be recommended for general use (see p. 477
of Ref. [1]).
Considering a mixed ®nite element discretization, in
whichever way the discretization has been formulated, it
is crucial that the scheme be analyzed for its mathe-
matical convergence characteristics [1,2]. The speci®c
conditions to be ful®lled are the consistency, ellipticity,
and inf±sup conditions. The consistency and ellipticity

0045-7949/01/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 0 0 ) 0 0 1 2 3 - 1
244 K.J. Bathe / Computers and Structures 79 (2001) 243±252

conditions are ful®lled relatively easily. However, the 2.1. The inf±sup condition using the complete bilinear
applicable inf±sup condition is frequently dicult to form
satisfy. Since an analytical proof to identify whether the
inf±sup condition is satis®ed is frequently out of reach, Consider a general variational problem given in a
numerical inf±sup test procedures have been designed Hilbert space U with a bilinear (symmetric or nonsym-
[1,3±6]. metric) form A /; w† de®ned on U  U. The ®rst ar-
It is also possible to design a mixed method in which gument in the bilinear form A
;
† is a solution function
an inf±sup condition is by-passed [7,8]. However, in that and the second argument is a weighting function. We
case, in general, numerical (arti®cial) constants enter the de®ne the following space:
formulation. The choice of these constants is a source of 
diculty, even though a mathematical analysis might ou
U ˆ uju 2 L2 Vol†; 2 L2 Vol†; k ˆ 1; 2; 3;
give guidelines as to the values to be used. It is more oxk

desirable to not introduce numerical constants in the
u ˆ 0 on Su ; 1†
formulation and design a method which does satisfy the
applicable inf±sup condition.
Our experience is that if the applicable inf±sup con- where we assume zero Dirichlet boundary conditions on
dition is satis®ed (and of course, the consistency and Su and L2 (Vol) is the space of square integrable func-
ellipticity requirements are ful®lled as well), the ®nite tions in the volume, ``Vol'', of the body considered,
element scheme is reliable and e€ective and never 
``fails''. This is the reason why we endeavor to only use L2 Vol† ˆ uju is defined in Vol and
®nite element discretization schemes that satisfy the Z 
condition. However, the inf±sup condition is a severe u2 dVol ˆ kuk2L2 Vol† < ‡ 1 : 2†
requirement and a method not ful®lling the condition Vol
might still be acceptable.
The objective of this paper is to present some latest Given a linear functional (F ; w) from U to R, we have
experiences in the use and numerical evaluation of the for the continuous problem:
inf±sup condition in various analysis areas. In the fol- Find / 2 U such that
lowing sections, we ®rst present the inf±sup condition in A /; w† ˆ F ; w† 8 w 2 U; 3†
two general forms, and then show how speci®c expres-
sions are applied to ®nite element formulations for in- where F is the forcing term. For nonzero Dirichlet
compressible analysis, acoustic analysis, plate and shell boundary conditions, the problem statement can of
structures and high Peclet and Reynolds number ¯ows. course also be reduced to Eq. (3) by including the e€ect
In the presentation, we endeavor to keep the discussion of the boundary conditions in F ; w† [1]. We assume that
short and hence do not give details of derivation but there exists a unique solution of Eq. (3).
refer to appropriate references, where these details can The ®nite dimensional subspace of U is de®ned as
be found. We conclude that mixed ®nite element for- follows:
mulations can now be evaluated as shown in this paper 
before it is claimed that an e€ective formulation has ouh
Uh ˆ uh juh 2 L2 Vol†; 2 L2 Vol†; k ˆ 1; 2; 3;
been reached. oxk
  
A brief explanation regarding the notation used
might be helpful. All continuous and ®nite element uh 2 Tn Vol m† ; uh ˆ 0 on Su ; 4†
variables are de®ned in RN ; N ˆ 1; 2 or 3 depending on  
the problem considered and are always simply presented where Tn Vol m† denotes a ®nite element interpolation
as italic symbols. In the de®nition of the functional of order n over the element m of volume Vol m† .
spaces used, the given properties apply to each vector The ®nite element solution of Eq. (3) is obtained by
component. However, the ®nite element matrices and solving the following ®nite dimensional problem:
the nodal point solution vectors are given in bold roman Find /h 2 Uh such that
symbols.
A /h ; w† ˆ F ; w† 8 w 2 Uh : 5†

To measure the quality of the ®nite element ap-

2. The inf±sup condition in a general setting
In this section we review the inf±sup condition in two
di€erent general forms. Applications are then presented
in Section 3.
proximation, we need to introduce norms. Let k
kS be
the norm that we use to measure the size of the solution
functions and k
kT be the norm for the weighting func-
tions. Then the continuity condition is
A g; w† 6 kA kgkS kwkT 8 g; w 2 U; 6†
 K.J. Bathe / Computers and Structures 79 (2001) 243±252 245

where kA is a (®nite) constant. This equation, assumed d /; Uh † ˆ inf k/ ÿ gh kS : 13†

gh 2Uh
satis®ed, simply states that the bilinear form A(g; w)
behaves normally. If the inf±sup condition is satis®ed, with c not too
Then, the classical condition to be satis®ed to have a small, and if kA is not too large, the constant (1 ‡ kA =c)
stable and optimal procedure for the solution of Eq. (5) in Eq. (8) is well behaved. Interpolation theory is next
is the inf±sup condition: used to obtain error bounds measured on the element
Inf±sup condition on A
;
†: size h. For example, consider the three-dimensional
$ a constant c > 0 and independent of crucial physical elasticity problem and let k
kS be the usual H1 Sobolev
parameters and the element size h such that norm. Interpolation theory gives that for problems with
A g; w† suciently smooth solutions, for some gh 2 Uh ,
inf sup P c: 7†
g2Uh w2Uh kgkS kwkT k/ ÿ gh kH 1 6 chk ; 14†

This condition ensures the solvability, stability and where h denotes the characteristic element size, k is the
optimality of the ®nite element solution. order of the complete polynomial in the ®nite elements
The inf±sup condition is crucial in establishing the used, and c is a constant, which depends on the problem
quality of the ®nite element approximation. In general, considered (the solution) but is independent of h. Com-
we have the following relation [1,2]: bining the relations (8) and (14), we have the optimal
  error bound
kA  
k/ ÿ /h kS 6 1 ‡ inf k/ ÿ gh kS : 8† kA k
c gh 2Uh k/ ÿ /h kH 1 6 c 1 ‡ h  c hk ; 15†
c
Let us show that the inf±sup condition implies the in-
equality (8). where c is a constant which depends on the problem
From the inequality (7) with gh ˆ vh ÿ /h , for any considered and is inversely proportional to c. Since c, kA
vh 2 Uh we have and c are independent of h, so is c .
The formulation that we considered in this section is
A vh ÿ /h ; wh † of very general mixed form. We now proceed to consider
ckvh ÿ /h kS 6 sup
wh 2Uh kwh kT a mixed formulation which is more specialized and is
A vh ÿ /; wh † ‡ A / ÿ /h ; wh † encountered widely in applications.
ˆ sup
wh 2Uh kwh kT
2.2. The inf±sup condition for the constraint mixed
A vh ÿ /; wh † formulation
ˆ sup
wh 2Uh kwh kT
kA kvh ÿ /kS kwh kT Consider now the general mixed formulation of a
6 sup
wh 2Uh kwh kT continuous problem in the classical form:
Given two Hilbert spaces U and E, ®nd u 2 U and
ˆ kA kvh ÿ /kS ; 9†
e 2 E such that
where we used Eq. (6) and that a u; v† ‡ b e; v† ˆ f ; v† 8 v 2 U;
16†
A / ÿ /h ; wh † ˆ 0 8 wh 2 Uh : 10† b g; u† ÿ c e; g† ˆ 0 8 g 2 E;

Using the triangle inequality, we thus have where a
;
†, b
;
† and c
;
† are bilinear forms, with
a
;
† and c
;
† symmetric forms, f ;
† is a linear form
k/h ÿ /kS 6 k/h ÿ vh kS ‡ kvh ÿ /kS and, of course, U and E are the spaces corresponding to
kA the problem considered. The second relation in Eq. (16)
6 kvh ÿ /kS ‡ kvh ÿ /kS
c provides of course the constraint in the ``mixed ®elds'' of
  the formulation. As in the previous discussion, for
kA
ˆ 1‡ kvh ÿ /kS ; 11† simplicity of presentation, we assume that the prescribed
c
boundary conditions on the solution variables are ho-
which proves the inequality (8). mogeneously zero. We also assume that
We may note that we can write Eq. (8) as a v; w† 6 ka k v kU k w kU 8 v; w 2 U;
17†
k/h ÿ /kS 6 b d /; Uh †; 12† b q; v† 6 kb k v kU k q kE 8 v 2 U; q 2 E;

where b is a constant, equal to (1 ‡ kA =c), and d /; Uh † where we use k
kU and k
kE to measure the size of the
is the distance between the exact solution / and the ®nite functions in U and E, and ka and kb are continuity
element interpolation space Uh , constants, and that
246 K.J. Bathe / Computers and Structures 79 (2001) 243±252
c q; q† 6 kc k q1 kE k q2 kE 8 q1 ; q2 2 E;
c q; q† P 0 8 q 2 E
with kc also a continuity constant. As in Section 2.1, we
assume that the continuous problem has a unique so-
lution.
The ®nite element solution of this problem is ob-
tained by choosing ®nite-dimensional subspaces of U
and E, which we call Uh and Eh , respectively. Then the
®nite element problem is:
Find uh 2 Uh and eh 2 Eh such that
a uh ; v† ‡ b eh ; v† ˆ f ; v† 8 v 2 Uh ;
b g; uh † ÿ c eh ; g† ˆ 0 8 g 2 Eh :
This mixed formulation should satisfy the following
two classical conditions to have an optimal procedure:
(i) Ellipticity of a
;
† on the null-space of b
;
†:
9 a constant a > 0 such that
a v; v† P akvk2U 8 v 2 Uh0 ;
where
Uh0 ˆ fv 2 Uh jb w; v† ˆ 0 8 w 2 Eh g
is the null space of b
;
†. Here we assume that c
;
† can
be zero, see the relations (18).
(ii) Inf±sup condition on b
;
†:
9 a constant c > 0 independent of crucial physical pa-
rameters and the element size h, such that
b g; v† b g; v†
sup P c sup 8 g 2 Eh :
v2Uh kvkU v2U kvkU
This inf±sup condition follows from the Banach theo-
rems [9] and corresponds to a general form. In speci®c
formulations, we seek to rewrite the supremum expres-
sion on the right-hand side to obtain a more tractable
form (see Sections 3.1±3.4).
When the inf±sup condition holds (and the ellipticity
condition is satis®ed), the following relation can be es-
tablished [1,2]:
ÿ where j is the bulk modulus,
ku ÿ uh kU ‡ ke ÿ eh kE 6 c inf ku ÿ vkU ‡ ke ÿ gkE †
v2Uh ; g2Eh
with c a constant which depends on ka , kb , kc , a and c
(and not h).
Assuming that the solution is suciently smooth and
using interpolation theory, the optimal error bounds are
obtained
ku ÿ uh kU ‡ ke ÿ eh kE  c hk ;
where k is the complete polynomial order used in the
interpolation space Uh , the interpolations in the space
Eh are performed with appropriate complementary or-
ders, and c is a (new) constant which depends on the
constant c in Eq. (23) and hence on c (actually on cÿ1 )
18† and also on the problem considered (the solution) but is
independent of the element size h. Note that Eq. (23) is
like Eq. (8) and Eq. (24) is like Eq. (15).
Hence, as in Section 2.1, it is crucial that the inf±sup
condition be satis®ed in order to have an e€ective ®nite
element procedure.
3. Applications of the inf±sup condition
The objective of this section is to show how the inf±
19† sup conditions in Eqs. (7) and (22) are applied and
evaluated in various analysis ®elds. In some cases ana-
lytical evaluations have been achieved (see, for example,
Ref. [1]) but we focus here on those ®nite element for-
mulations, where an analytical evaluation has been out
of reach. In that case, a numerical evaluation is very
valuable, and while not as encompassing as an analytical
20† proof, our experience is that when the numerical test is
passed, in fact, the inf±sup condition is satis®ed. A nu-
merical test is like the patch test (used for noncon-
21† forming displacement discretizations) usually performed
numerically instead of analytically.
3.1. Incompressible elasticity (and Stokes ¯ow)
For the solution of an incompressible (or almost in-
compressible) elasticity problem, the displacement/pres-
sure (u/p) ®nite element discretizations are e€ective [1].
22† Let Vh be the ®nite element displacement interpolation
space and Qh be the ®nite element pressure interpolation
space (corresponding to the spaces V and Q of the
continuous problem). Then the ®nite element problem
is:
Find uh 2 Vh and ph 2 Qh such that
aIE uh ; vh † ‡ b ph ; vh † ˆ f ; vh † 8 vh 2 Vh ;
25†
1
b qh ; uh † ÿ ph ; qh † ˆ 0 8 qh 2 Qh ;
j
Z
23† b qh; vh † ˆ qh div vh dVol 26†
Vol
and aIE
;
† is the bilinear form of the deviatoric strain
energy.
Considering the presentation in Section 2.2, the
norms used here are k
kU  k
kH 1 and k
kE  k
kL2 . The
inf±sup condition is in this case obtained from Eq. (22)
24† by bounding the supremum on the right-hand side as
R
qh div v dVol
c1 kqh kL2 P sup Vol P c2 kqh kL2 27†
v2V kvkH 1
with c1 and c2 constants.
 K.J. Bathe / Computers and Structures 79 (2001) 243±252 247

Hence, the L2 -norm of qh is equivalent to the right-

hand side of Eq. (22) and the inf±sup condition is for
this problem
R
qh div vh dVol
inf sup Vol ˆ ch P c > 0 28†
qh 2Qh vh 2Vh kvh kH 1 kqh kL2

with c a constant independent of h. Note that we con-

sider here the almost incompressible case and the fully
incompressible (the limit) case. This inf±sup condition is
frequently referred to as the Babuska±Brezzi (or simply
BB) condition because of the seminal papers by Babuska
[10] and Brezzi [11]. The condition is also referred to as
the LBB condition including the name of Ladyzhens-
kaya [12]. Various u/p ®nite element discretization
schemes have been proven analytically to satisfy this
condition (see, for example Ref. [1]). However, some
element discretizations and discretizations based on
distorted element meshes (used virtually always in en-
gineering practice) could not (yet) be proven analytically
to satisfy the condition, and therefore a numerical test
was proposed by Chapelle and Bathe [3].
For the inf±sup test, the matrices G h and S h are de-
®ned for a given ®nite element discretization. The mesh
used for a given geometry, including the displacement
boundary conditions, gives
Z modes). Hence, the element is optimal as discussed in
qh div vh dVol ˆ W Th G h V h ; 29†
Vol
kqh k2L2 ˆ W Th G h W h ; 30†
kvh k2H 1 ˆ V Th S h V h ; 31†
where W h and V h are the ®nite element nodal point
solution vectors. In practice, the semi-norms instead
of the full norms are used (based on the Poincare±
Friedrichs inequality [1]).
The smallest nonzero eigenvalue kk of the following
problem is then calculated to obtain the inf±sup value ch ,
as detailed in Ref. [1],
G h / ˆ k S h /: 32†
p
The value kk is equal to the inf±sup value ch in Eq. (28)
for the discretization considered provided that there are
no spurious pressure modes. The number of spurious
pressure modes can be predicted from
kpm ˆ k ÿ nu ÿ np ‡ 1†; 33†
where nu is the number of displacement degrees of
freedom and np is the number of pressure degrees of
freedom. If kpm > 0, the ®nite element discretization
contains the constant pressure mode or spurious pres-
sure modes [1]. The physical constant pressure mode can
be eliminated by changing the boundary conditions. If a
spurious pressure mode is present, ch is zero and the inf±
Fig. 1. Inf±sup test results for the (almost or fully) incom-
pressible analysis case. A square cantilever plate using meshes
of N by N elements is considered.
sup test is of course not passed (the inf±sup condition is
not satis®ed).
The test involves the calculation of ch for a sequence
of meshes with increasingly ®ner discretizations; typi-
cally meshes are used with element sizes h; h=2;
h=4; h=8; . . . If the inf±sup values for these discretiza-
tions do not show a decrease towards zero, the test is
passed (provided that there are also no spurious pressure
Section 2.2. Since the constant ``c'' in Eq. (28) does not
depend on h (and of course not on the bulk modulus
which approaches in®nity as PoissonÕs ratio approaches
0.5), the element does not ``lock''.
Fig. 1 shows results reported in Ref. [3] for the testing
of various elements. Note that any loading and material
properties do not enter the test. The boundary condi-
tions, on the other hand, do enter the test and hence a
judiciously chosen problem should be considered.
Note that this test also shows whether an element is
optimal in almost incompressible conditions such as
encountered in inelastic analysis. Indeed, only elements
that pass this test can be recommended for the solution
of almost or fully incompressible media, including, in
particular, also elasto-plastic conditions [1].
The de®nition of G h in Eq. (30) is for the u/p for-
mulation in which inter-element discontinuous pressure
interpolations are used. In case inter-element continuous
pressure interpolations are employed (referred to as the
u/p±c formulation) the matrix G h is di€erent [1].
The test is also directly applicable to ¯uid ¯ow ele-
ments used to solve Stokes ¯ow by simply using the
velocity instead of the displacement variables, and can
of course also be employed for elements based on en-
hanced strain ®elds [13].
3.2. Acoustic ¯uid
If an acoustic ¯uid is modeled using the displacements
and pressure as variables, the formulation discussed in
248 K.J. Bathe / Computers and Structures 79 (2001) 243±252

Find uh ˆ bh ; wh † 2 Vh and ch 2 Gh ˆ Rh Vh † such

that 1
aRM bh ; gh † ‡ ch ; Rh rnh ÿ gh †† ˆ g; nh †
 2 
t
8 vh ˆ gh ; nh † 2 Vh ; ch ÿ Rh rwh ÿ bh †; 1h ˆ 0
lk
8 1h 2 Gh ; 36†

where aRM
;
† is the bilinear form corresponding to the

bending action and g represents the transverse load,
both normalized by t3 , t is the thickness of the plate, l
is the shear modulus and k is the shear correction factor.
Fig. 2. Inf±sup test results for the irrotational acoustic ¯uid The reduction operator Rh is introduced to ``weaken''
elements. Fluid in a rigid cavity using meshes of N by N ele- the shear strain constraint in order to not have shear
ments is considered [15].
locking in the formulation. The operator must be chosen
judiciously to preserve consistency and to satisfy the
ellipticity and inf±sup conditions [1,4]. Note that the
formulation in Eq. (36) represents a quite general solu-
the previous section is obtained. In this case, the bilinear
tion approach.
form aIE
;
† is zero (or of small magnitude) because the
The inf±sup condition is in this case obtained from
¯uid is (almost) inviscid, and many physical zero (or
Eq. (22) by using the following equivalence for the su-
small) frequencies are encountered. Some of these zero
premum on the right-hand side:
frequencies can be eliminated by imposing the constraint
of zero vorticity using [14,15] 1h ; v† 1h ; Rh rn ÿ g†† 1 ; v†
c1 sup P sup P c2 sup h ;
v2C kvkC vˆ g;n†2V kvkV v2C kvkC
k
ruˆ ; 34† 37†
a

where k is a vorticity moment and a is a constant of large where c1 are c2 are constants and C corresponds to the
continuous space
value.

This constraint is an additional one to the constraint C ˆ v; v 2 L2 Vol†; rot v 2 L2 Vol†; v
s ˆ 0 on S
on the pressure included already, see Eq. (25), and the
resulting inf±sup expression is then with s the unit tangent vector on the boundary surface
R R S. Hence, the inf±sup condition is [4]
qh div vh dVol ‡ Vol kh rot vh dVol
inf sup Vol ÿ
1h ; Rh rnh ÿ gh ††
qh 2Qh ; kh 2Kh vh 2Vh kvh kH 1 kqh kL2 ‡ kkh kL2 inf sup ˆ ch P c > 0 38†
1h 2Gh vh 2Vh kvh kV k1h kC0
ˆ ch P c > 0;
35† with c independent of h, and

where Kh is the discrete space of vorticity moments, 1h ; v†

kh 2 Kh , and c is to be a constant independent of h. The
inf±sup test is performed as described in Section 3.1 and
Fig. 2 shows some results obtained in testing acoustic
¯uid elements [15].
3.3. Reissner±Mindlin plate bending
The solution of plate bending problems is accom-
plished e€ectively using elements based on Reissner±
Mindlin plate theory [1]. Let Vh ˆ Bh  Wh be the ®nite
element interpolation space of the plate section rotations
bh (in Bh ) and transverse displacement wh (in Wh ), with
V being the space of the continuous problem, and let Gh
be the ®nite element interpolation space of the transverse
shear strains times lk=t2 . Then the ®nite element prob-
lem is [4]:
k1h kC0 ˆ sup : 39†
v2C kvkC
The diculty in evaluating the inf±sup condition lies in
that the C0 -norm involves the complete continuous
space. Therefore, the inf±sup test for Eq. (38) consists of
two parts. In the ®rst part, the inf±sup expression cor-
responding to C0h is evaluated.
Consider a given ®nite element discretization. Using
the particular discretization scheme for the transverse
shear strains, section rotations and transverse displace-
ment, we have
1
Here ch is the vector of plate bending shear strains times
lk=t2 . Since ch 2 Gh is only used in this equation, a confusion
with the inf±sup value used in Eq. (38) is ruled out.
 K.J. Bathe / Computers and Structures 79 (2001) 243±252
kvh k2C ˆ kvh k2L2 ‡ L2 k rot vh k2L2
ˆ W Th G h ‡ Qh †W h ˆ W Th Dh W h
with L the characteristic length of the plate problem
considered, and
kvh k2V ˆ kgh k2H 1 ‡ knh k2H 1 ˆ V Th S h V h :
In Eqs. (40) and (41), the vectors W h and V h contain
the ®nite element nodal displacement and rotation
variables. The inf±sup value ch is given by [4]
q
ch P ch ˆ dhmin =rhmax ; 42†
where dhmin stands for the smallest nonzero eigenvalue of
Dh and rhmax is the largest eigenvalue of Sh . In practice,
the semi-norms instead of the full norms are used in Eq.
(41) (based on the Poincare±Friedrichs inequality [1]).
In the second part, we need to show that the C0h - and
0
C -norms are equivalent so that in fact the conclusions
reached using Eq. (38) with the C0h -norm are also ap-
plicable when the C0 -norm is used. Considering a se-
quence of meshes by subdividing the element size h
consecutively, we have
1h ; vh †
k1h kC0 ˆ sup ! k1h kC0 :
h=n vh 2Ch=n kvh kCh=n n!1
So that if we consider
k1h kC0
hn h† ˆ inf h
; 44†
1h 2Ch k1h kC0
h=n
we can ®nd an estimate for the constant (when n is large)
k1h kC0
h1 h† ˆ inf h
: 45†
1h 2Ch k1h kC0
Then if
h1 h† P h > 0
for some h independent of h, we can conclude that the
C0h - and C0 -norms are equivalent. In some cases, it is
possible to prove analytically that h2 h† is a lower bound
on h1 h† [16] but a general numerical procedure is also
given in Refs. [4,16].
An alternative way to proceed is to consider a plate
element to be a special case of a shell element and use the
inf±sup test considered in the next section.
We should note that the formulation given in Eq.
(36) is quite general, but the e€ectiveness of a discreti-
zation depends of course on the speci®c interpolation
spaces chosen and the reduction operator Rh . The choice
of the function spaces for the MITC family of plate
bending elements, which satisfy the consistency, ellip-
ticity and inf±sup conditions, is discussed in Refs. [1,17]
and Fig. 3 shows some inf±sup test results for a few
249
40†
41†
Fig. 3. Inf±sup test results for some MITC plate-bending ele-
ments. A clamped square plate using meshes of N by N dis-
torted elements is considered [4].
MITC plate bending elements. (Note that the values of
ch , see Eq. (42), are plotted here).
3.4. General shell conditions
An e€ective shell ®nite element must be applicable to
membrane- and bending-dominated shell problems.
43† Ideally, the element shows optimal convergence char-
acteristics in both problem areas. Elements formulated
using pure displacement interpolations are e€ective in
membrane-dominated situations but ``lock'' when bend-
ing is encountered. To arrive at a general shell ®nite el-
ement discretization, a mixed formulation must be used.
The formulation must preserve consistency for all strain
terms, satisfy the ellipticity condition and, considering
the bending-dominated case, satisfy the inf±sup condi-
tion.
For the analysis of general shell problems, the MITC
shell elements are quite e€ective [1,18]. Let Uh be the
displacement interpolation space (including the dis-
46† placements and section rotations) and Eh be the strain
interpolation space; then the MITC shell element for-
mulation considering the bending-dominated case can
be written in the classical form of Eq. (16):
Find Uh 2 Uh and Eh 2 Eh such that
ÿ AS

A e Uh †; eAS Vh † ‡ B Eh ; Vh † ˆ Fb ; Vh † 8 Vh 2 Uh ;
B wh ; Uh † ÿ t2 C Eh ; wh † ˆ 0 8 wh 2 Eh ;
47†
where A
;
† is the bilinear form containing the bending
strains eAS obtained by tying to the displacements, Eh
corresponds to the membrane and shear strain compo-
nents divided by t2 , t is the shell thickness, B
;
† is the
bilinear form corresponding to the membrane and shear
actions, C
;
† is the bilinear form constraining with
B
;
† the strain interpolations to the displacements, and
Fb ;
† is the linear form of the loads divided by t3 . Since
250 K.J. Bathe / Computers and Structures 79 (2001) 243±252

the loading is scaled by t3 , the bending-dominated case is

considered.
The inf±sup condition employed is [5]
B wh ; Vh †
inf sup ˆ ch P c > 0; 48†
wh 2Eh Vh 2Uh kVh kU kwh kL2

where we used that

B w; V † 6 ckwkL2 kV kU 8 w; V † 2 L2  U 49†

and hence for Eq. (22)

B w; V †
sup 6 ckwkL2 ; 50†
V 2U kV kU

where c is a constant.
The bilinear form B
;
† evaluates the membrane and
shear strain energies and these can be calculated intro-
ducing
ÿ

e h V h ˆ C eAS Wh †; eAS Vh † ;
W Th K 51†

where K e h is in practice obtained by simply integrating

the usual strain terms but only on the mid-surface of the
shell (that is, using one-point integration in the element
thickness direction).
The matrix S h is used to calculate the norm on Vh ,
Fig. 4. Inf±sup test results for MITC shell elements. A partly
kVh k2U ˆ V Th S h V h : 52† clamped hyperbolic paraboloid shell using uniform meshes is
considered [5]: (a) Quadrilateral elements and (b) triangular
Instead p the L2 -norm
of 
 on wh , we use the equivalent elements.
expression C wh ; wh †, and then the inf±sup value ch in
Eq. (48) can be evaluated by calculating the smallest
nonzero eigenvalue kk of the problem Find hh 2 Uh such that
Ke h / ˆ kS h /: 53† a hh ; wh † ˆ b; wh † 8 wh 2 Uh : 54†
p
The value kk is the inf±sup value ch in Eq. (48) for the This problem is of the kind discussed in Section 2.1,
discretization considered. The inf±sup test is performed where we note that the bilinear form a
;
† is nonsym-
as described in Section 3.1. In the test, of course, a metric. The inf±sup condition was given as
bending-dominated problem needs to be considered, and
a nh ; wh †
according to our experience a hyperbolic paraboloid inf sup ˆ ch P c > 0: 55†
shell provides a suitable problem [5]. This same test can nh 2Uh w 2Uh
h
knh kS kwh kT
of course also be applied to Reissner±Mindlin plate
bending elements and to Timoshenko beam elements. If For the evaluation of the inf±sup value we de®ne the
the test is passed, the element is e€ective for bending- matrices Ah , Sh and Th . Considering a given discretiza-
dominated problems. However, if the test is not passed, tion we have, using Eq. (54) [6]
we cannot conclude that the element does not satisfy the Ah x ˆ b 56†
inf±sup condition in Eq. (22) because the conditions in
Eqs. (22) and (48) are not equivalent (see Eq. (50)). and the norms
Fig. 4 shows some inf±sup test results calculated for
MITC shell elements [5]. knh k2S ˆ nTh S h nh ; 57†

3.5. High P
eclet number ¯ow kwh k2T ˆ wTh T h wh ; 58†

Consider a high Peclet number ¯ow problem, in where nh and wh are the nodal values corresponding to nh
which the velocity is prescribed so that the only un- and wh , respectively. Then, as derived in Ref. [6], the inf±
known is the temperature, h. The ®nite element problem sup value ch is obtained by solving the following eigen-
is [1]: value problem for the smallest eigenvalue kmin ,
 K.J. Bathe / Computers and Structures 79 (2001) 243±252
Fig. 5. Inf±sup test results for one-dimensional high Peclet
number ¯ow problem [6]. The convection±di€usion problem
corresponds to a constant prescribed velocity throughout the
domain and prescribed temperatures at both ends of the do-
main. 16 equal-sized elements are used.
ÿ

ATh T ÿ1
h Ah / ˆ k S h /: 59†
p
The inf±sup value ch is equal to kmin and in general
depends on the Peclet number (Pe). The inf±sup test is
performed by calculating ch for a given discretization as
the Peclet number is increased to very large values. If the
inf±sup value does not decrease, but stabilizes to remain
larger than a value greater than zero, the test is passed.
Alternatively, we can consider a problem of very high
Peclet number, use a ®ne discretization and then calcu-
late ch for increasingly coarser meshes. The test is passed
if ch does not decrease but stabilizes above a value larger
than zero.
These tests were performed in Ref. [6] for various
upwind schemes applied to a one-dimensional test
problem (Fig. 5). The curves given in Fig. 5 show that,
as expected, the standard Galerkin method is unstable
whereas the other methods are stable. The inf±sup test
results also show the amount of arti®cial di€usion con-
tained in the method; that is, the smaller the inf±sup
values, the less arti®cial di€usion is contained in the
method.
4. Concluding remarks
The objective of this paper was to present some latest
experiences in the use and evaluation of the inf±sup
condition for mixed ®nite element formulations. This
condition is frequently the crucial condition to ensure
the stability and optimality of a discretization scheme.
However, it can be dicult, or impossible, to evaluate
analytically whether the condition is satis®ed. For this
251
reason, numerical inf±sup tests have been designed. As
shown in this paper these tests are applied in a straight-
forward manner (like the patch test for displacement-
based incompatible models) to discretizations applicable
to various analysis ®elds. Inf±sup tests for analysis ®elds
not considered herein can be similarly performed.
Our experiences are that when the numerical inf±sup
test is passed, the inf±sup condition is satis®ed. Hence,
mixed ®nite element formulations can now be evaluated
as shown in this paper before it is claimed that an ef-
fective formulation has been developed.
Acknowledgements
I would like to thank F. Brezzi, University of Pavia
and D. Chapelle, INRIA for their long and fruitful co-
operation with me on the theory and methods discussed
in this paper.
References
[1] Bathe KJ. Finite element procedures. Englewood Cli€s,
NJ: Prentice Hall; 1996.
[2] Brezzi F, Bathe KJ. A discourse on the stability conditions
for mixed ®nite element formulations. Comp Meth Appl
Mech Engng 1990;82:27±57.
[3] Chapelle D, Bathe KJ. The inf-sup test. Comput Struct
1993;47:537±45.
[4] Iosilevich A, Bathe KJ, Brezzi F. On evaluating the inf-sup
condition for plate bending elements. Int J Num Meth
Engng 1997;40:3639±63.
[5] Bathe KJ, Iosilevich A, Chapelle D. An inf-sup test for
shell ®nite elements. Comput Struct 2000;75:439±56.
[6] Bathe KJ, Hendriana D, Brezzi F, Sangalli G. Inf-sup
testing of upwind methods. Int J Num Meth Engng
2000;48:745±60.
[7] Brezzi F, Pitk aranta J. On the stabilization of ®nite
element approximations of the Stokes equations. In:
Hackbush W, editor, Ecient solutions of elliptic systems.
Notes on numerical ¯uid mechanics, vol. 10, p. 11±9.
Vieweg Braunschweig, Wiesbaden, 1984.
[8] Hughes TJR, Franca LP, Balestre M. A new ®nite element
formulation for computational ¯uid dynamics: V. Cir-
cumventing the Babuska±Brezzi condition: A stable
Petrov±Galerkin formulation of the Stokes problem ac-
commodating equal-order interpolations. Comp Meth
Appl Mech Engng 1986;59:85±99.
[9] Banach S. Theorie des Operations Lineaires. Warszawa,
1932.
[10] Babuska I. The ®nite element method with Lagrangian
multipliers. Numerische Mathematik 1973;20:179±92.
[11] Brezzi F. On the existence, uniqueness and approximation
of saddle-point problems arising from Lagrangian multi-
pliers. Revue francßaise dÕautomatique informatique recher-
che operationnelle. Analyse Numerique 1974;8:129±51.
[12] Ladyzhenskaya OA. The mathematical theory of viscous
incompressible ¯ow. New York: Gordon and Breach; 1969.
252 K.J. Bathe / Computers and Structures 79 (2001) 243±252
[13] Pantuso D, Bathe KJ. A four-node quadrilateral mixed-
interpolated element for solids and ¯uids. Math Models
Meth Appl Sci 1995;5:1113±28.
[14] Wang X, Bathe KJ. Displacement/pressure based mixed ®-
nite element formulations for acoustic ¯uid-structure inter-
action problems. Int J Num Meth Engng 1997;40:2001±17.
[15] Wang X, Bathe KJ. On mixed elements for acoustic ¯uid-
structure interactions. Math Models Meth Appl Sci 1997;
7:329±43.
[16] Iosilevich A, Bathe KJ, Brezzi F. Numerical inf±sup
analysis of MITC plate bending elements. Proceedings,
1996 American Mathematical Society Seminar on Plates
and Shells. Universite Laval, Canada, 1996.
[17] Bathe KJ, Bucalem ML, Brezzi F. Displacement and stress
convergence of the MITC plate bending elements. J Engng
Computat 1990;7:291±302.
[18] Bathe KJ, Iosilevich A, Chapelle D. An evaluation of the
MITC shell elements. Comput Struct 2000;75:1±30.